548 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py2; RUN: opt < %s -passes=instcombine -mtriple=x86_64-unknown-unknown -S | FileCheck %s3 4; Verify that instcombine is able to fold identity shuffles.5 6define <16 x i8> @identity_test(<16 x i8> %InVec) {7; CHECK-LABEL: @identity_test(8; CHECK-NEXT: ret <16 x i8> [[INVEC:%.*]]9;10 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>)11 ret <16 x i8> %112}13 14define <32 x i8> @identity_test_avx2(<32 x i8> %InVec) {15; CHECK-LABEL: @identity_test_avx2(16; CHECK-NEXT: ret <32 x i8> [[INVEC:%.*]]17;18 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>)19 ret <32 x i8> %120}21 22define <64 x i8> @identity_test_avx512(<64 x i8> %InVec) {23; CHECK-LABEL: @identity_test_avx512(24; CHECK-NEXT: ret <64 x i8> [[INVEC:%.*]]25;26 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>)27 ret <64 x i8> %128}29 30; Verify that instcombine is able to fold byte shuffles with zero masks.31 32define <16 x i8> @fold_to_zero_vector(<16 x i8> %InVec) {33; CHECK-LABEL: @fold_to_zero_vector(34; CHECK-NEXT: ret <16 x i8> zeroinitializer35;36 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)37 ret <16 x i8> %138}39 40define <32 x i8> @fold_to_zero_vector_avx2(<32 x i8> %InVec) {41; CHECK-LABEL: @fold_to_zero_vector_avx2(42; CHECK-NEXT: ret <32 x i8> zeroinitializer43;44 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)45 ret <32 x i8> %146}47 48define <64 x i8> @fold_to_zero_vector_avx512(<64 x i8> %InVec) {49; CHECK-LABEL: @fold_to_zero_vector_avx512(50; CHECK-NEXT: ret <64 x i8> zeroinitializer51;52 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)53 ret <64 x i8> %154}55 56; Instcombine should be able to fold the following byte shuffle to a builtin shufflevector57; with a shuffle mask of all zeroes.58 59define <16 x i8> @splat_test(<16 x i8> %InVec) {60; CHECK-LABEL: @splat_test(61; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> poison, <16 x i32> zeroinitializer62; CHECK-NEXT: ret <16 x i8> [[TMP1]]63;64 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> zeroinitializer)65 ret <16 x i8> %166}67 68; In the test case below, elements in the low 128-bit lane of the result69; vector are equal to the lower byte of %InVec (shuffle index 0).70; Elements in the high 128-bit lane of the result vector are equal to71; the lower byte in the high 128-bit lane of %InVec (shuffle index 16).72 73define <32 x i8> @splat_test_avx2(<32 x i8> %InVec) {74; CHECK-LABEL: @splat_test_avx2(75; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> poison, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>76; CHECK-NEXT: ret <32 x i8> [[TMP1]]77;78 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> zeroinitializer)79 ret <32 x i8> %180}81 82define <64 x i8> @splat_test_avx512(<64 x i8> %InVec) {83; CHECK-LABEL: @splat_test_avx512(84; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> poison, <64 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48>85; CHECK-NEXT: ret <64 x i8> [[TMP1]]86;87 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> zeroinitializer)88 ret <64 x i8> %189}90 91; Each of the byte shuffles in the following tests is equivalent to a blend between92; vector %InVec and a vector of all zeroes.93 94define <16 x i8> @blend1(<16 x i8> %InVec) {95; CHECK-LABEL: @blend1(96; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <16 x i32> <i32 16, i32 1, i32 16, i32 3, i32 16, i32 5, i32 16, i32 7, i32 16, i32 9, i32 16, i32 11, i32 16, i32 13, i32 16, i32 15>97; CHECK-NEXT: ret <16 x i8> [[TMP1]]98;99 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 -128, i8 1, i8 -128, i8 3, i8 -128, i8 5, i8 -128, i8 7, i8 -128, i8 9, i8 -128, i8 11, i8 -128, i8 13, i8 -128, i8 15>)100 ret <16 x i8> %1101}102 103define <16 x i8> @blend2(<16 x i8> %InVec) {104; CHECK-LABEL: @blend2(105; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <16 x i32> <i32 16, i32 16, i32 2, i32 3, i32 16, i32 16, i32 6, i32 7, i32 16, i32 16, i32 10, i32 11, i32 16, i32 16, i32 14, i32 15>106; CHECK-NEXT: ret <16 x i8> [[TMP1]]107;108 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 -128, i8 -128, i8 2, i8 3, i8 -128, i8 -128, i8 6, i8 7, i8 -128, i8 -128, i8 10, i8 11, i8 -128, i8 -128, i8 14, i8 15>)109 ret <16 x i8> %1110}111 112define <16 x i8> @blend3(<16 x i8> %InVec) {113; CHECK-LABEL: @blend3(114; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 4, i32 5, i32 6, i32 7, i32 16, i32 16, i32 16, i32 16, i32 12, i32 13, i32 14, i32 15>115; CHECK-NEXT: ret <16 x i8> [[TMP1]]116;117 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 12, i8 13, i8 14, i8 15>)118 ret <16 x i8> %1119}120 121define <16 x i8> @blend4(<16 x i8> %InVec) {122; CHECK-LABEL: @blend4(123; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>124; CHECK-NEXT: ret <16 x i8> [[TMP1]]125;126 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>)127 ret <16 x i8> %1128}129 130define <16 x i8> @blend5(<16 x i8> %InVec) {131; CHECK-LABEL: @blend5(132; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>133; CHECK-NEXT: ret <16 x i8> [[TMP1]]134;135 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)136 ret <16 x i8> %1137}138 139define <16 x i8> @blend6(<16 x i8> %InVec) {140; CHECK-LABEL: @blend6(141; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <16 x i32> <i32 0, i32 1, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>142; CHECK-NEXT: ret <16 x i8> [[TMP1]]143;144 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 0, i8 1, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)145 ret <16 x i8> %1146}147 148define <32 x i8> @blend1_avx2(<32 x i8> %InVec) {149; CHECK-LABEL: @blend1_avx2(150; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <32 x i32> <i32 32, i32 1, i32 32, i32 3, i32 32, i32 5, i32 32, i32 7, i32 32, i32 9, i32 32, i32 11, i32 32, i32 13, i32 32, i32 15, i32 48, i32 17, i32 48, i32 19, i32 48, i32 21, i32 48, i32 23, i32 48, i32 25, i32 48, i32 27, i32 48, i32 29, i32 48, i32 31>151; CHECK-NEXT: ret <32 x i8> [[TMP1]]152;153 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 -128, i8 1, i8 -128, i8 3, i8 -128, i8 5, i8 -128, i8 7, i8 -128, i8 9, i8 -128, i8 11, i8 -128, i8 13, i8 -128, i8 15, i8 -128, i8 1, i8 -128, i8 3, i8 -128, i8 5, i8 -128, i8 7, i8 -128, i8 9, i8 -128, i8 11, i8 -128, i8 13, i8 -128, i8 15>)154 ret <32 x i8> %1155}156 157define <32 x i8> @blend2_avx2(<32 x i8> %InVec) {158; CHECK-LABEL: @blend2_avx2(159; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <32 x i32> <i32 32, i32 32, i32 2, i32 3, i32 32, i32 32, i32 6, i32 7, i32 32, i32 32, i32 10, i32 11, i32 32, i32 32, i32 14, i32 15, i32 48, i32 48, i32 18, i32 19, i32 48, i32 48, i32 22, i32 23, i32 48, i32 48, i32 26, i32 27, i32 48, i32 48, i32 30, i32 31>160; CHECK-NEXT: ret <32 x i8> [[TMP1]]161;162 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 -128, i8 -128, i8 2, i8 3, i8 -128, i8 -128, i8 6, i8 7, i8 -128, i8 -128, i8 10, i8 11, i8 -128, i8 -128, i8 14, i8 15, i8 -128, i8 -128, i8 2, i8 3, i8 -128, i8 -128, i8 6, i8 7, i8 -128, i8 -128, i8 10, i8 11, i8 -128, i8 -128, i8 14, i8 15>)163 ret <32 x i8> %1164}165 166define <32 x i8> @blend3_avx2(<32 x i8> %InVec) {167; CHECK-LABEL: @blend3_avx2(168; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <32 x i32> <i32 32, i32 32, i32 32, i32 32, i32 4, i32 5, i32 6, i32 7, i32 32, i32 32, i32 32, i32 32, i32 12, i32 13, i32 14, i32 15, i32 48, i32 48, i32 48, i32 48, i32 20, i32 21, i32 22, i32 23, i32 48, i32 48, i32 48, i32 48, i32 28, i32 29, i32 30, i32 31>169; CHECK-NEXT: ret <32 x i8> [[TMP1]]170;171 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 12, i8 13, i8 14, i8 15, i8 -128, i8 -128, i8 -128, i8 -128, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 12, i8 13, i8 14, i8 15>)172 ret <32 x i8> %1173}174 175define <32 x i8> @blend4_avx2(<32 x i8> %InVec) {176; CHECK-LABEL: @blend4_avx2(177; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <32 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>178; CHECK-NEXT: ret <32 x i8> [[TMP1]]179;180 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>)181 ret <32 x i8> %1182}183 184define <32 x i8> @blend5_avx2(<32 x i8> %InVec) {185; CHECK-LABEL: @blend5_avx2(186; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 16, i32 17, i32 18, i32 19, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48>187; CHECK-NEXT: ret <32 x i8> [[TMP1]]188;189 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 0, i8 1, i8 2, i8 3, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)190 ret <32 x i8> %1191}192 193define <32 x i8> @blend6_avx2(<32 x i8> %InVec) {194; CHECK-LABEL: @blend6_avx2(195; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <32 x i32> <i32 0, i32 1, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 16, i32 17, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48>196; CHECK-NEXT: ret <32 x i8> [[TMP1]]197;198 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 0, i8 1, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 0, i8 1, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)199 ret <32 x i8> %1200}201 202define <64 x i8> @blend1_avx512(<64 x i8> %InVec) {203; CHECK-LABEL: @blend1_avx512(204; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <64 x i32> <i32 64, i32 1, i32 64, i32 3, i32 64, i32 5, i32 64, i32 7, i32 64, i32 9, i32 64, i32 11, i32 64, i32 13, i32 64, i32 15, i32 80, i32 17, i32 80, i32 19, i32 80, i32 21, i32 80, i32 23, i32 80, i32 25, i32 80, i32 27, i32 80, i32 29, i32 80, i32 31, i32 96, i32 33, i32 96, i32 35, i32 96, i32 37, i32 96, i32 39, i32 96, i32 41, i32 96, i32 43, i32 96, i32 45, i32 96, i32 47, i32 112, i32 49, i32 112, i32 51, i32 112, i32 53, i32 112, i32 55, i32 112, i32 57, i32 112, i32 59, i32 112, i32 61, i32 112, i32 63>205; CHECK-NEXT: ret <64 x i8> [[TMP1]]206;207 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 -128, i8 1, i8 -128, i8 3, i8 -128, i8 5, i8 -128, i8 7, i8 -128, i8 9, i8 -128, i8 11, i8 -128, i8 13, i8 -128, i8 15, i8 -128, i8 1, i8 -128, i8 3, i8 -128, i8 5, i8 -128, i8 7, i8 -128, i8 9, i8 -128, i8 11, i8 -128, i8 13, i8 -128, i8 15, i8 -128, i8 1, i8 -128, i8 3, i8 -128, i8 5, i8 -128, i8 7, i8 -128, i8 9, i8 -128, i8 11, i8 -128, i8 13, i8 -128, i8 15, i8 -128, i8 1, i8 -128, i8 3, i8 -128, i8 5, i8 -128, i8 7, i8 -128, i8 9, i8 -128, i8 11, i8 -128, i8 13, i8 -128, i8 15>)208 ret <64 x i8> %1209}210 211define <64 x i8> @blend2_avx512(<64 x i8> %InVec) {212; CHECK-LABEL: @blend2_avx512(213; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <64 x i32> <i32 64, i32 64, i32 2, i32 3, i32 64, i32 64, i32 6, i32 7, i32 64, i32 64, i32 10, i32 11, i32 64, i32 64, i32 14, i32 15, i32 80, i32 80, i32 18, i32 19, i32 80, i32 80, i32 22, i32 23, i32 80, i32 80, i32 26, i32 27, i32 80, i32 80, i32 30, i32 31, i32 96, i32 96, i32 34, i32 35, i32 96, i32 96, i32 38, i32 39, i32 96, i32 96, i32 42, i32 43, i32 96, i32 96, i32 46, i32 47, i32 112, i32 112, i32 50, i32 51, i32 112, i32 112, i32 54, i32 55, i32 112, i32 112, i32 58, i32 59, i32 112, i32 112, i32 62, i32 63>214; CHECK-NEXT: ret <64 x i8> [[TMP1]]215;216 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 -128, i8 -128, i8 2, i8 3, i8 -128, i8 -128, i8 6, i8 7, i8 -128, i8 -128, i8 10, i8 11, i8 -128, i8 -128, i8 14, i8 15, i8 -128, i8 -128, i8 2, i8 3, i8 -128, i8 -128, i8 6, i8 7, i8 -128, i8 -128, i8 10, i8 11, i8 -128, i8 -128, i8 14, i8 15, i8 -128, i8 -128, i8 2, i8 3, i8 -128, i8 -128, i8 6, i8 7, i8 -128, i8 -128, i8 10, i8 11, i8 -128, i8 -128, i8 14, i8 15, i8 -128, i8 -128, i8 2, i8 3, i8 -128, i8 -128, i8 6, i8 7, i8 -128, i8 -128, i8 10, i8 11, i8 -128, i8 -128, i8 14, i8 15>)217 ret <64 x i8> %1218}219 220define <64 x i8> @blend3_avx512(<64 x i8> %InVec) {221; CHECK-LABEL: @blend3_avx512(222; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <64 x i32> <i32 64, i32 64, i32 64, i32 64, i32 4, i32 5, i32 6, i32 7, i32 64, i32 64, i32 64, i32 64, i32 12, i32 13, i32 14, i32 15, i32 80, i32 80, i32 80, i32 80, i32 20, i32 21, i32 22, i32 23, i32 80, i32 80, i32 80, i32 80, i32 28, i32 29, i32 30, i32 31, i32 96, i32 96, i32 96, i32 96, i32 36, i32 37, i32 38, i32 39, i32 96, i32 96, i32 96, i32 96, i32 44, i32 45, i32 46, i32 47, i32 112, i32 112, i32 112, i32 112, i32 52, i32 53, i32 54, i32 55, i32 112, i32 112, i32 112, i32 112, i32 60, i32 61, i32 62, i32 63>223; CHECK-NEXT: ret <64 x i8> [[TMP1]]224;225 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 12, i8 13, i8 14, i8 15, i8 -128, i8 -128, i8 -128, i8 -128, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 12, i8 13, i8 14, i8 15, i8 -128, i8 -128, i8 -128, i8 -128, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 12, i8 13, i8 14, i8 15, i8 -128, i8 -128, i8 -128, i8 -128, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 12, i8 13, i8 14, i8 15>)226 ret <64 x i8> %1227}228 229define <64 x i8> @blend4_avx512(<64 x i8> %InVec) {230; CHECK-LABEL: @blend4_avx512(231; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <64 x i32> <i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>232; CHECK-NEXT: ret <64 x i8> [[TMP1]]233;234 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>)235 ret <64 x i8> %1236}237 238define <64 x i8> @blend5_avx512(<64 x i8> %InVec) {239; CHECK-LABEL: @blend5_avx512(240; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 16, i32 17, i32 18, i32 19, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 32, i32 33, i32 34, i32 35, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 48, i32 49, i32 50, i32 51, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112>241; CHECK-NEXT: ret <64 x i8> [[TMP1]]242;243 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 0, i8 1, i8 2, i8 3, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 0, i8 1, i8 2, i8 3, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 0, i8 1, i8 2, i8 3, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)244 ret <64 x i8> %1245}246 247define <64 x i8> @blend6_avx512(<64 x i8> %InVec) {248; CHECK-LABEL: @blend6_avx512(249; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <64 x i32> <i32 0, i32 1, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 16, i32 17, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 32, i32 33, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 48, i32 49, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112>250; CHECK-NEXT: ret <64 x i8> [[TMP1]]251;252 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 0, i8 1, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128,i8 0, i8 1, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 0, i8 1, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 0, i8 1, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)253 ret <64 x i8> %1254}255 256; movq idiom.257define <16 x i8> @movq_idiom(<16 x i8> %InVec) {258; CHECK-LABEL: @movq_idiom(259; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>260; CHECK-NEXT: ret <16 x i8> [[TMP1]]261;262 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)263 ret <16 x i8> %1264}265 266define <32 x i8> @movq_idiom_avx2(<32 x i8> %InVec) {267; CHECK-LABEL: @movq_idiom_avx2(268; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48, i32 48>269; CHECK-NEXT: ret <32 x i8> [[TMP1]]270;271 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)272 ret <32 x i8> %1273}274 275define <64 x i8> @movq_idiom_avx512(<64 x i8> %InVec) {276; CHECK-LABEL: @movq_idiom_avx512(277; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 80, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 96, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112, i32 112>278; CHECK-NEXT: ret <64 x i8> [[TMP1]]279;280 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128, i8 -128>)281 ret <64 x i8> %1282}283 284; Vector permutations using byte shuffles.285 286define <16 x i8> @permute1(<16 x i8> %InVec) {287; CHECK-LABEL: @permute1(288; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> poison, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15, i32 12, i32 13, i32 14, i32 15>289; CHECK-NEXT: ret <16 x i8> [[TMP1]]290;291 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 4, i8 5, i8 6, i8 7, i8 4, i8 5, i8 6, i8 7, i8 12, i8 13, i8 14, i8 15, i8 12, i8 13, i8 14, i8 15>)292 ret <16 x i8> %1293}294 295define <16 x i8> @permute2(<16 x i8> %InVec) {296; CHECK-LABEL: @permute2(297; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>298; CHECK-NEXT: ret <16 x i8> [[TMP1]]299;300 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>)301 ret <16 x i8> %1302}303 304define <32 x i8> @permute1_avx2(<32 x i8> %InVec) {305; CHECK-LABEL: @permute1_avx2(306; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> poison, <32 x i32> <i32 4, i32 5, i32 6, i32 7, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15, i32 12, i32 13, i32 14, i32 15, i32 20, i32 21, i32 22, i32 23, i32 20, i32 21, i32 22, i32 23, i32 28, i32 29, i32 30, i32 31, i32 28, i32 29, i32 30, i32 31>307; CHECK-NEXT: ret <32 x i8> [[TMP1]]308;309 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 4, i8 5, i8 6, i8 7, i8 4, i8 5, i8 6, i8 7, i8 12, i8 13, i8 14, i8 15, i8 12, i8 13, i8 14, i8 15, i8 4, i8 5, i8 6, i8 7, i8 4, i8 5, i8 6, i8 7, i8 12, i8 13, i8 14, i8 15, i8 12, i8 13, i8 14, i8 15>)310 ret <32 x i8> %1311}312 313define <32 x i8> @permute2_avx2(<32 x i8> %InVec) {314; CHECK-LABEL: @permute2_avx2(315; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>316; CHECK-NEXT: ret <32 x i8> [[TMP1]]317;318 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>)319 ret <32 x i8> %1320}321 322define <64 x i8> @permute1_avx512(<64 x i8> %InVec) {323; CHECK-LABEL: @permute1_avx512(324; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> poison, <64 x i32> <i32 4, i32 5, i32 6, i32 7, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15, i32 12, i32 13, i32 14, i32 15, i32 20, i32 21, i32 22, i32 23, i32 20, i32 21, i32 22, i32 23, i32 28, i32 29, i32 30, i32 31, i32 28, i32 29, i32 30, i32 31, i32 36, i32 37, i32 38, i32 39, i32 36, i32 37, i32 38, i32 39, i32 44, i32 45, i32 46, i32 47, i32 44, i32 45, i32 46, i32 47, i32 52, i32 53, i32 54, i32 55, i32 52, i32 53, i32 54, i32 55, i32 60, i32 61, i32 62, i32 63, i32 60, i32 61, i32 62, i32 63>325; CHECK-NEXT: ret <64 x i8> [[TMP1]]326;327 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 4, i8 5, i8 6, i8 7, i8 4, i8 5, i8 6, i8 7, i8 12, i8 13, i8 14, i8 15, i8 12, i8 13, i8 14, i8 15, i8 4, i8 5, i8 6, i8 7, i8 4, i8 5, i8 6, i8 7, i8 12, i8 13, i8 14, i8 15, i8 12, i8 13, i8 14, i8 15, i8 4, i8 5, i8 6, i8 7, i8 4, i8 5, i8 6, i8 7, i8 12, i8 13, i8 14, i8 15, i8 12, i8 13, i8 14, i8 15, i8 4, i8 5, i8 6, i8 7, i8 4, i8 5, i8 6, i8 7, i8 12, i8 13, i8 14, i8 15, i8 12, i8 13, i8 14, i8 15>)328 ret <64 x i8> %1329}330 331define <64 x i8> @permute2_avx512(<64 x i8> %InVec) {332; CHECK-LABEL: @permute2_avx512(333; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>334; CHECK-NEXT: ret <64 x i8> [[TMP1]]335;336 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>)337 ret <64 x i8> %1338}339 340; Test that instcombine correctly folds a pshufb with values that341; are not -128 and that are not encoded in four bits.342 343define <16 x i8> @identity_test2_2(<16 x i8> %InVec) {344; CHECK-LABEL: @identity_test2_2(345; CHECK-NEXT: ret <16 x i8> [[INVEC:%.*]]346;347 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23, i8 24, i8 25, i8 26, i8 27, i8 28, i8 29, i8 30, i8 31>)348 ret <16 x i8> %1349}350 351define <32 x i8> @identity_test_avx2_2(<32 x i8> %InVec) {352; CHECK-LABEL: @identity_test_avx2_2(353; CHECK-NEXT: ret <32 x i8> [[INVEC:%.*]]354;355 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 16, i8 33, i8 66, i8 19, i8 36, i8 69, i8 22, i8 39, i8 72, i8 25, i8 42, i8 75, i8 28, i8 45, i8 78, i8 31, i8 48, i8 81, i8 34, i8 51, i8 84, i8 37, i8 54, i8 87, i8 40, i8 57, i8 90, i8 43, i8 60, i8 93, i8 46, i8 63>)356 ret <32 x i8> %1357}358 359define <64 x i8> @identity_test_avx512_2(<64 x i8> %InVec) {360; CHECK-LABEL: @identity_test_avx512_2(361; CHECK-NEXT: ret <64 x i8> [[INVEC:%.*]]362;363 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 16, i8 33, i8 66, i8 19, i8 36, i8 69, i8 22, i8 39, i8 72, i8 25, i8 42, i8 75, i8 28, i8 45, i8 78, i8 31, i8 48, i8 81, i8 34, i8 51, i8 84, i8 37, i8 54, i8 87, i8 40, i8 57, i8 90, i8 43, i8 60, i8 93, i8 46, i8 63, i8 96, i8 49, i8 66, i8 99, i8 52, i8 69, i8 102, i8 55, i8 72, i8 105, i8 58, i8 75, i8 108, i8 61, i8 78, i8 111, i8 64, i8 81, i8 114, i8 67, i8 84, i8 117, i8 70, i8 87, i8 120, i8 73, i8 90, i8 123, i8 76, i8 93, i8 126, i8 79>)364 ret <64 x i8> %1365}366 367define <16 x i8> @fold_to_zero_vector_2(<16 x i8> %InVec) {368; CHECK-LABEL: @fold_to_zero_vector_2(369; CHECK-NEXT: ret <16 x i8> zeroinitializer370;371 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 -125, i8 -1, i8 -53, i8 -32, i8 -4, i8 -7, i8 -33, i8 -66, i8 -99, i8 -120, i8 -100, i8 -22, i8 -17, i8 -1, i8 -11, i8 -15>)372 ret <16 x i8> %1373}374 375define <32 x i8> @fold_to_zero_vector_avx2_2(<32 x i8> %InVec) {376; CHECK-LABEL: @fold_to_zero_vector_avx2_2(377; CHECK-NEXT: ret <32 x i8> zeroinitializer378;379 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 -127, i8 -1, i8 -53, i8 -32, i8 -4, i8 -7, i8 -33, i8 -66, i8 -99, i8 -120, i8 -100, i8 -22, i8 -17, i8 -1, i8 -11, i8 -15, i8 -126, i8 -2, i8 -52, i8 -31, i8 -5, i8 -8, i8 -34, i8 -67, i8 -100, i8 -119, i8 -101, i8 -23, i8 -16, i8 -2, i8 -12, i8 -16>)380 ret <32 x i8> %1381}382 383define <64 x i8> @fold_to_zero_vector_avx512_2(<64 x i8> %InVec) {384; CHECK-LABEL: @fold_to_zero_vector_avx512_2(385; CHECK-NEXT: ret <64 x i8> zeroinitializer386;387 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 -127, i8 -1, i8 -53, i8 -32, i8 -4, i8 -7, i8 -33, i8 -66, i8 -99, i8 -120, i8 -100, i8 -22, i8 -17, i8 -1, i8 -11, i8 -15, i8 -126, i8 -2, i8 -52, i8 -31, i8 -5, i8 -8, i8 -34, i8 -67, i8 -100, i8 -119, i8 -101, i8 -23, i8 -16, i8 -2, i8 -12, i8 -16, i8 -125, i8 -3, i8 -51, i8 -30, i8 -6, i8 -9, i8 -35, i8 -68, i8 -101, i8 -118, i8 -102, i8 -24, i8 -15, i8 -3, i8 -13, i8 -17, i8 -124, i8 -4, i8 -56, i8 -29, i8 -7, i8 -10, i8 -36, i8 -69, i8 -102, i8 -117, i8 -103, i8 -25, i8 -14, i8 -4, i8 -14, i8 -18>)388 ret <64 x i8> %1389}390 391define <16 x i8> @permute3(<16 x i8> %InVec) {392; CHECK-LABEL: @permute3(393; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>394; CHECK-NEXT: ret <16 x i8> [[TMP1]]395;396 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 48, i8 17, i8 34, i8 51, i8 20, i8 37, i8 54, i8 23, i8 16, i8 49, i8 66, i8 19, i8 52, i8 69, i8 22, i8 55>)397 ret <16 x i8> %1398}399 400define <32 x i8> @permute3_avx2(<32 x i8> %InVec) {401; CHECK-LABEL: @permute3_avx2(402; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> poison, <32 x i32> <i32 4, i32 5, i32 6, i32 7, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15, i32 12, i32 13, i32 14, i32 15, i32 20, i32 21, i32 22, i32 23, i32 20, i32 21, i32 22, i32 23, i32 28, i32 29, i32 30, i32 31, i32 28, i32 29, i32 30, i32 31>403; CHECK-NEXT: ret <32 x i8> [[TMP1]]404;405 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 52, i8 21, i8 38, i8 55, i8 20, i8 37, i8 54, i8 23, i8 28, i8 61, i8 78, i8 31, i8 60, i8 29, i8 30, i8 79, i8 52, i8 21, i8 38, i8 55, i8 20, i8 53, i8 102, i8 23, i8 92, i8 93, i8 94, i8 95, i8 108, i8 109, i8 110, i8 111>)406 ret <32 x i8> %1407}408 409define <64 x i8> @permute3_avx512(<64 x i8> %InVec) {410; CHECK-LABEL: @permute3_avx512(411; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> poison, <64 x i32> <i32 4, i32 5, i32 6, i32 7, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15, i32 12, i32 13, i32 14, i32 15, i32 20, i32 21, i32 22, i32 23, i32 20, i32 21, i32 22, i32 23, i32 28, i32 29, i32 30, i32 31, i32 28, i32 29, i32 30, i32 31, i32 36, i32 37, i32 38, i32 39, i32 36, i32 37, i32 38, i32 39, i32 44, i32 45, i32 46, i32 47, i32 44, i32 45, i32 46, i32 47, i32 52, i32 53, i32 54, i32 55, i32 52, i32 53, i32 54, i32 55, i32 60, i32 61, i32 62, i32 63, i32 60, i32 61, i32 62, i32 63>412; CHECK-NEXT: ret <64 x i8> [[TMP1]]413;414 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 52, i8 21, i8 38, i8 55, i8 20, i8 37, i8 54, i8 23, i8 28, i8 61, i8 78, i8 31, i8 60, i8 29, i8 30, i8 79, i8 52, i8 21, i8 38, i8 55, i8 20, i8 53, i8 102, i8 23, i8 92, i8 93, i8 94, i8 95, i8 108, i8 109, i8 110, i8 111, i8 52, i8 21, i8 38, i8 55, i8 20, i8 37, i8 54, i8 23, i8 28, i8 61, i8 78, i8 31, i8 60, i8 29, i8 30, i8 79, i8 52, i8 21, i8 38, i8 55, i8 20, i8 53, i8 102, i8 23, i8 108, i8 109, i8 110, i8 111, i8 124, i8 125, i8 126, i8 127>)415 ret <64 x i8> %1416}417 418; FIXME: Verify that instcombine is able to fold constant byte shuffles with poison mask elements.419 420define <16 x i8> @fold_with_poison_elts(<16 x i8> %InVec) {421; CHECK-LABEL: @fold_with_poison_elts(422; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <16 x i8> [[INVEC:%.*]], <16 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <16 x i32> <i32 0, i32 16, i32 poison, i32 16, i32 1, i32 16, i32 poison, i32 16, i32 2, i32 16, i32 poison, i32 16, i32 3, i32 16, i32 poison, i32 16>423; CHECK-NEXT: ret <16 x i8> [[TMP1]]424;425 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> <i8 0, i8 -128, i8 poison, i8 -128, i8 1, i8 -128, i8 poison, i8 -128, i8 2, i8 -128, i8 poison, i8 -128, i8 3, i8 -128, i8 poison, i8 -128>)426 ret <16 x i8> %1427}428 429define <32 x i8> @fold_with_poison_elts_avx2(<32 x i8> %InVec) {430; CHECK-LABEL: @fold_with_poison_elts_avx2(431; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <32 x i8> [[INVEC:%.*]], <32 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <32 x i32> <i32 0, i32 32, i32 poison, i32 32, i32 1, i32 32, i32 poison, i32 32, i32 2, i32 32, i32 poison, i32 32, i32 3, i32 32, i32 poison, i32 32, i32 16, i32 48, i32 poison, i32 48, i32 17, i32 48, i32 poison, i32 48, i32 18, i32 48, i32 poison, i32 48, i32 19, i32 48, i32 poison, i32 48>432; CHECK-NEXT: ret <32 x i8> [[TMP1]]433;434 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> <i8 0, i8 -128, i8 poison, i8 -128, i8 1, i8 -128, i8 poison, i8 -128, i8 2, i8 -128, i8 poison, i8 -128, i8 3, i8 -128, i8 poison, i8 -128, i8 0, i8 -128, i8 poison, i8 -128, i8 1, i8 -128, i8 poison, i8 -128, i8 2, i8 -128, i8 poison, i8 -128, i8 3, i8 -128, i8 poison, i8 -128>)435 ret <32 x i8> %1436}437 438define <64 x i8> @fold_with_poison_elts_avx512(<64 x i8> %InVec) {439; CHECK-LABEL: @fold_with_poison_elts_avx512(440; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <64 x i8> [[INVEC:%.*]], <64 x i8> <i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>, <64 x i32> <i32 0, i32 64, i32 poison, i32 64, i32 1, i32 64, i32 poison, i32 64, i32 2, i32 64, i32 poison, i32 64, i32 3, i32 64, i32 poison, i32 64, i32 16, i32 80, i32 poison, i32 80, i32 17, i32 80, i32 poison, i32 80, i32 18, i32 80, i32 poison, i32 80, i32 19, i32 80, i32 poison, i32 80, i32 32, i32 96, i32 poison, i32 96, i32 33, i32 96, i32 poison, i32 96, i32 34, i32 96, i32 poison, i32 96, i32 35, i32 96, i32 poison, i32 96, i32 48, i32 112, i32 poison, i32 112, i32 49, i32 112, i32 poison, i32 112, i32 50, i32 112, i32 poison, i32 112, i32 51, i32 112, i32 poison, i32 112>441; CHECK-NEXT: ret <64 x i8> [[TMP1]]442;443 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> <i8 0, i8 -128, i8 poison, i8 -128, i8 1, i8 -128, i8 poison, i8 -128, i8 2, i8 -128, i8 poison, i8 -128, i8 3, i8 -128, i8 poison, i8 -128, i8 0, i8 -128, i8 poison, i8 -128, i8 1, i8 -128, i8 poison, i8 -128, i8 2, i8 -128, i8 poison, i8 -128, i8 3, i8 -128, i8 poison, i8 -128, i8 0, i8 -128, i8 poison, i8 -128, i8 1, i8 -128, i8 poison, i8 -128, i8 2, i8 -128, i8 poison, i8 -128, i8 3, i8 -128, i8 poison, i8 -128, i8 0, i8 -128, i8 poison, i8 -128, i8 1, i8 -128, i8 poison, i8 -128, i8 2, i8 -128, i8 poison, i8 -128, i8 3, i8 -128, i8 poison, i8 -128>)444 ret <64 x i8> %1445}446 447define <16 x i8> @fold_with_allpoison_elts(<16 x i8> %InVec) {448; CHECK-LABEL: @fold_with_allpoison_elts(449; CHECK-NEXT: ret <16 x i8> poison450;451 %1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> poison)452 ret <16 x i8> %1453}454 455define <32 x i8> @fold_with_allpoison_elts_avx2(<32 x i8> %InVec) {456; CHECK-LABEL: @fold_with_allpoison_elts_avx2(457; CHECK-NEXT: ret <32 x i8> poison458;459 %1 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> poison)460 ret <32 x i8> %1461}462 463define <64 x i8> @fold_with_allpoison_elts_avx512(<64 x i8> %InVec) {464; CHECK-LABEL: @fold_with_allpoison_elts_avx512(465; CHECK-NEXT: ret <64 x i8> poison466;467 %1 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> poison)468 ret <64 x i8> %1469}470 471; Demanded bits tests (PR106256)472 473define <16 x i8> @demanded_bits_mask(<16 x i8> %InVec, <16 x i8> %InMask) {474; CHECK-LABEL: @demanded_bits_mask(475; CHECK-NEXT: [[S:%.*]] = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> [[INVEC:%.*]], <16 x i8> [[INMASK:%.*]])476; CHECK-NEXT: ret <16 x i8> [[S]]477;478 %m = or <16 x i8> %InMask, <i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112>479 %s = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> %m)480 ret <16 x i8> %s481}482 483define <32 x i8> @demanded_bits_mask_avx2(<32 x i8> %InVec, <32 x i8> %InMask) {484; CHECK-LABEL: @demanded_bits_mask_avx2(485; CHECK-NEXT: [[S:%.*]] = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> [[INVEC:%.*]], <32 x i8> [[INMASK:%.*]])486; CHECK-NEXT: ret <32 x i8> [[S]]487;488 %m = or <32 x i8> %InMask, <i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112>489 %s = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> %m)490 ret <32 x i8> %s491}492 493define <64 x i8> @demanded_bits_mask_avx512(<64 x i8> %InVec, <64 x i8> %InMask) {494; CHECK-LABEL: @demanded_bits_mask_avx512(495; CHECK-NEXT: [[S:%.*]] = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> [[INVEC:%.*]], <64 x i8> [[INMASK:%.*]])496; CHECK-NEXT: ret <64 x i8> [[S]]497;498 %m = or <64 x i8> %InMask, <i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112, i8 16, i8 48, i8 112, i8 112>499 %s = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> %m)500 ret <64 x i8> %s501}502 503; Demanded elts tests.504 505define <16 x i8> @demanded_elts_insertion(<16 x i8> %InVec, <16 x i8> %BaseMask, i8 %M0, i8 %M15) {506; CHECK-LABEL: @demanded_elts_insertion(507; CHECK-NEXT: [[TMP1:%.*]] = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> [[INVEC:%.*]], <16 x i8> [[BASEMASK:%.*]])508; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> poison, <16 x i32> <i32 poison, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 poison>509; CHECK-NEXT: ret <16 x i8> [[TMP2]]510;511 %1 = insertelement <16 x i8> %BaseMask, i8 %M0, i32 0512 %2 = insertelement <16 x i8> %1, i8 %M15, i32 15513 %3 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %InVec, <16 x i8> %2)514 %4 = shufflevector <16 x i8> %3, <16 x i8> poison, <16 x i32> <i32 poison, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 poison>515 ret <16 x i8> %4516}517 518define <32 x i8> @demanded_elts_insertion_avx2(<32 x i8> %InVec, <32 x i8> %BaseMask, i8 %M0, i8 %M22) {519; CHECK-LABEL: @demanded_elts_insertion_avx2(520; CHECK-NEXT: [[TMP1:%.*]] = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> [[INVEC:%.*]], <32 x i8> [[BASEMASK:%.*]])521; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <32 x i8> [[TMP1]], <32 x i8> poison, <32 x i32> <i32 poison, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 poison, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>522; CHECK-NEXT: ret <32 x i8> [[TMP2]]523;524 %1 = insertelement <32 x i8> %BaseMask, i8 %M0, i32 0525 %2 = insertelement <32 x i8> %1, i8 %M22, i32 22526 %3 = tail call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %InVec, <32 x i8> %2)527 %4 = shufflevector <32 x i8> %3, <32 x i8> poison, <32 x i32> <i32 poison, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 poison, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>528 ret <32 x i8> %4529}530 531define <64 x i8> @demanded_elts_insertion_avx512(<64 x i8> %InVec, <64 x i8> %BaseMask, i8 %M0, i8 %M30) {532; CHECK-LABEL: @demanded_elts_insertion_avx512(533; CHECK-NEXT: [[TMP1:%.*]] = insertelement <64 x i8> poison, i8 [[M0:%.*]], i64 0534; CHECK-NEXT: [[TMP2:%.*]] = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> [[INVEC:%.*]], <64 x i8> [[TMP1]])535; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <64 x i8> [[TMP2]], <64 x i8> poison, <64 x i32> zeroinitializer536; CHECK-NEXT: ret <64 x i8> [[TMP3]]537;538 %1 = insertelement <64 x i8> %BaseMask, i8 %M0, i32 0539 %2 = insertelement <64 x i8> %1, i8 %M30, i32 30540 %3 = tail call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %InVec, <64 x i8> %2)541 %4 = shufflevector <64 x i8> %3, <64 x i8> poison, <64 x i32> zeroinitializer542 ret <64 x i8> %4543}544 545declare <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8>, <16 x i8>)546declare <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8>, <32 x i8>)547declare <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8>, <64 x i8>)548